xref: /freebsd/sys/amd64/include/vmm.h (revision 864c53ead899f7838cd2e1cca3b485a4a82f5cdc)
1 /*-
2  * Copyright (c) 2011 NetApp, Inc.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  *
14  * THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND
15  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17  * ARE DISCLAIMED.  IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE
18  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24  * SUCH DAMAGE.
25  *
26  * $FreeBSD$
27  */
28 
29 #ifndef _VMM_H_
30 #define	_VMM_H_
31 
32 enum vm_suspend_how {
33 	VM_SUSPEND_NONE,
34 	VM_SUSPEND_RESET,
35 	VM_SUSPEND_POWEROFF,
36 	VM_SUSPEND_HALT,
37 	VM_SUSPEND_LAST
38 };
39 
40 #ifdef _KERNEL
41 
42 #define	VM_MAX_NAMELEN	32
43 
44 struct vm;
45 struct vm_exception;
46 struct vm_memory_segment;
47 struct seg_desc;
48 struct vm_exit;
49 struct vm_run;
50 struct vhpet;
51 struct vioapic;
52 struct vlapic;
53 struct vmspace;
54 struct vm_object;
55 struct pmap;
56 
57 enum vm_reg_name;
58 enum x2apic_state;
59 
60 typedef int	(*vmm_init_func_t)(int ipinum);
61 typedef int	(*vmm_cleanup_func_t)(void);
62 typedef void	(*vmm_resume_func_t)(void);
63 typedef void *	(*vmi_init_func_t)(struct vm *vm, struct pmap *pmap);
64 typedef int	(*vmi_run_func_t)(void *vmi, int vcpu, register_t rip,
65 				  struct pmap *pmap, void *rendezvous_cookie,
66 				  void *suspend_cookie);
67 typedef void	(*vmi_cleanup_func_t)(void *vmi);
68 typedef int	(*vmi_get_register_t)(void *vmi, int vcpu, int num,
69 				      uint64_t *retval);
70 typedef int	(*vmi_set_register_t)(void *vmi, int vcpu, int num,
71 				      uint64_t val);
72 typedef int	(*vmi_get_desc_t)(void *vmi, int vcpu, int num,
73 				  struct seg_desc *desc);
74 typedef int	(*vmi_set_desc_t)(void *vmi, int vcpu, int num,
75 				  struct seg_desc *desc);
76 typedef int	(*vmi_get_cap_t)(void *vmi, int vcpu, int num, int *retval);
77 typedef int	(*vmi_set_cap_t)(void *vmi, int vcpu, int num, int val);
78 typedef struct vmspace * (*vmi_vmspace_alloc)(vm_offset_t min, vm_offset_t max);
79 typedef void	(*vmi_vmspace_free)(struct vmspace *vmspace);
80 typedef struct vlapic * (*vmi_vlapic_init)(void *vmi, int vcpu);
81 typedef void	(*vmi_vlapic_cleanup)(void *vmi, struct vlapic *vlapic);
82 
83 struct vmm_ops {
84 	vmm_init_func_t		init;		/* module wide initialization */
85 	vmm_cleanup_func_t	cleanup;
86 	vmm_resume_func_t	resume;
87 
88 	vmi_init_func_t		vminit;		/* vm-specific initialization */
89 	vmi_run_func_t		vmrun;
90 	vmi_cleanup_func_t	vmcleanup;
91 	vmi_get_register_t	vmgetreg;
92 	vmi_set_register_t	vmsetreg;
93 	vmi_get_desc_t		vmgetdesc;
94 	vmi_set_desc_t		vmsetdesc;
95 	vmi_get_cap_t		vmgetcap;
96 	vmi_set_cap_t		vmsetcap;
97 	vmi_vmspace_alloc	vmspace_alloc;
98 	vmi_vmspace_free	vmspace_free;
99 	vmi_vlapic_init		vlapic_init;
100 	vmi_vlapic_cleanup	vlapic_cleanup;
101 };
102 
103 extern struct vmm_ops vmm_ops_intel;
104 extern struct vmm_ops vmm_ops_amd;
105 
106 int vm_create(const char *name, struct vm **retvm);
107 void vm_destroy(struct vm *vm);
108 int vm_reinit(struct vm *vm);
109 const char *vm_name(struct vm *vm);
110 int vm_malloc(struct vm *vm, vm_paddr_t gpa, size_t len);
111 int vm_map_mmio(struct vm *vm, vm_paddr_t gpa, size_t len, vm_paddr_t hpa);
112 int vm_unmap_mmio(struct vm *vm, vm_paddr_t gpa, size_t len);
113 void *vm_gpa_hold(struct vm *, vm_paddr_t gpa, size_t len, int prot,
114 		  void **cookie);
115 void vm_gpa_release(void *cookie);
116 int vm_gpabase2memseg(struct vm *vm, vm_paddr_t gpabase,
117 	      struct vm_memory_segment *seg);
118 int vm_get_memobj(struct vm *vm, vm_paddr_t gpa, size_t len,
119 		  vm_offset_t *offset, struct vm_object **object);
120 boolean_t vm_mem_allocated(struct vm *vm, vm_paddr_t gpa);
121 int vm_get_register(struct vm *vm, int vcpu, int reg, uint64_t *retval);
122 int vm_set_register(struct vm *vm, int vcpu, int reg, uint64_t val);
123 int vm_get_seg_desc(struct vm *vm, int vcpu, int reg,
124 		    struct seg_desc *ret_desc);
125 int vm_set_seg_desc(struct vm *vm, int vcpu, int reg,
126 		    struct seg_desc *desc);
127 int vm_run(struct vm *vm, struct vm_run *vmrun);
128 int vm_suspend(struct vm *vm, enum vm_suspend_how how);
129 int vm_inject_nmi(struct vm *vm, int vcpu);
130 int vm_nmi_pending(struct vm *vm, int vcpuid);
131 void vm_nmi_clear(struct vm *vm, int vcpuid);
132 int vm_inject_extint(struct vm *vm, int vcpu);
133 int vm_extint_pending(struct vm *vm, int vcpuid);
134 void vm_extint_clear(struct vm *vm, int vcpuid);
135 uint64_t *vm_guest_msrs(struct vm *vm, int cpu);
136 struct vlapic *vm_lapic(struct vm *vm, int cpu);
137 struct vioapic *vm_ioapic(struct vm *vm);
138 struct vhpet *vm_hpet(struct vm *vm);
139 int vm_get_capability(struct vm *vm, int vcpu, int type, int *val);
140 int vm_set_capability(struct vm *vm, int vcpu, int type, int val);
141 int vm_get_x2apic_state(struct vm *vm, int vcpu, enum x2apic_state *state);
142 int vm_set_x2apic_state(struct vm *vm, int vcpu, enum x2apic_state state);
143 int vm_apicid2vcpuid(struct vm *vm, int apicid);
144 int vm_activate_cpu(struct vm *vm, int vcpu);
145 cpuset_t vm_active_cpus(struct vm *vm);
146 cpuset_t vm_suspended_cpus(struct vm *vm);
147 struct vm_exit *vm_exitinfo(struct vm *vm, int vcpuid);
148 void vm_exit_suspended(struct vm *vm, int vcpuid, uint64_t rip);
149 
150 /*
151  * Rendezvous all vcpus specified in 'dest' and execute 'func(arg)'.
152  * The rendezvous 'func(arg)' is not allowed to do anything that will
153  * cause the thread to be put to sleep.
154  *
155  * If the rendezvous is being initiated from a vcpu context then the
156  * 'vcpuid' must refer to that vcpu, otherwise it should be set to -1.
157  *
158  * The caller cannot hold any locks when initiating the rendezvous.
159  *
160  * The implementation of this API may cause vcpus other than those specified
161  * by 'dest' to be stalled. The caller should not rely on any vcpus making
162  * forward progress when the rendezvous is in progress.
163  */
164 typedef void (*vm_rendezvous_func_t)(struct vm *vm, int vcpuid, void *arg);
165 void vm_smp_rendezvous(struct vm *vm, int vcpuid, cpuset_t dest,
166     vm_rendezvous_func_t func, void *arg);
167 
168 static __inline int
169 vcpu_rendezvous_pending(void *rendezvous_cookie)
170 {
171 
172 	return (*(uintptr_t *)rendezvous_cookie != 0);
173 }
174 
175 static __inline int
176 vcpu_suspended(void *suspend_cookie)
177 {
178 
179 	return (*(int *)suspend_cookie);
180 }
181 
182 /*
183  * Return 1 if device indicated by bus/slot/func is supposed to be a
184  * pci passthrough device.
185  *
186  * Return 0 otherwise.
187  */
188 int vmm_is_pptdev(int bus, int slot, int func);
189 
190 void *vm_iommu_domain(struct vm *vm);
191 
192 enum vcpu_state {
193 	VCPU_IDLE,
194 	VCPU_FROZEN,
195 	VCPU_RUNNING,
196 	VCPU_SLEEPING,
197 };
198 
199 int vcpu_set_state(struct vm *vm, int vcpu, enum vcpu_state state,
200     bool from_idle);
201 enum vcpu_state vcpu_get_state(struct vm *vm, int vcpu, int *hostcpu);
202 
203 static int __inline
204 vcpu_is_running(struct vm *vm, int vcpu, int *hostcpu)
205 {
206 	return (vcpu_get_state(vm, vcpu, hostcpu) == VCPU_RUNNING);
207 }
208 
209 void *vcpu_stats(struct vm *vm, int vcpu);
210 void vcpu_notify_event(struct vm *vm, int vcpuid, bool lapic_intr);
211 struct vmspace *vm_get_vmspace(struct vm *vm);
212 int vm_assign_pptdev(struct vm *vm, int bus, int slot, int func);
213 int vm_unassign_pptdev(struct vm *vm, int bus, int slot, int func);
214 struct vatpic *vm_atpic(struct vm *vm);
215 struct vatpit *vm_atpit(struct vm *vm);
216 
217 /*
218  * Inject exception 'vme' into the guest vcpu. This function returns 0 on
219  * success and non-zero on failure.
220  *
221  * Wrapper functions like 'vm_inject_gp()' should be preferred to calling
222  * this function directly because they enforce the trap-like or fault-like
223  * behavior of an exception.
224  *
225  * This function should only be called in the context of the thread that is
226  * executing this vcpu.
227  */
228 int vm_inject_exception(struct vm *vm, int vcpuid, struct vm_exception *vme);
229 
230 /*
231  * Returns 0 if there is no exception pending for this vcpu. Returns 1 if an
232  * exception is pending and also updates 'vme'. The pending exception is
233  * cleared when this function returns.
234  *
235  * This function should only be called in the context of the thread that is
236  * executing this vcpu.
237  */
238 int vm_exception_pending(struct vm *vm, int vcpuid, struct vm_exception *vme);
239 
240 void vm_inject_gp(struct vm *vm, int vcpuid); /* general protection fault */
241 void vm_inject_ud(struct vm *vm, int vcpuid); /* undefined instruction fault */
242 void vm_inject_pf(struct vm *vm, int vcpuid, int error_code, uint64_t cr2);
243 
244 enum vm_reg_name vm_segment_name(int seg_encoding);
245 
246 #endif	/* KERNEL */
247 
248 #define	VM_MAXCPU	16			/* maximum virtual cpus */
249 
250 /*
251  * Identifiers for architecturally defined registers.
252  */
253 enum vm_reg_name {
254 	VM_REG_GUEST_RAX,
255 	VM_REG_GUEST_RBX,
256 	VM_REG_GUEST_RCX,
257 	VM_REG_GUEST_RDX,
258 	VM_REG_GUEST_RSI,
259 	VM_REG_GUEST_RDI,
260 	VM_REG_GUEST_RBP,
261 	VM_REG_GUEST_R8,
262 	VM_REG_GUEST_R9,
263 	VM_REG_GUEST_R10,
264 	VM_REG_GUEST_R11,
265 	VM_REG_GUEST_R12,
266 	VM_REG_GUEST_R13,
267 	VM_REG_GUEST_R14,
268 	VM_REG_GUEST_R15,
269 	VM_REG_GUEST_CR0,
270 	VM_REG_GUEST_CR3,
271 	VM_REG_GUEST_CR4,
272 	VM_REG_GUEST_DR7,
273 	VM_REG_GUEST_RSP,
274 	VM_REG_GUEST_RIP,
275 	VM_REG_GUEST_RFLAGS,
276 	VM_REG_GUEST_ES,
277 	VM_REG_GUEST_CS,
278 	VM_REG_GUEST_SS,
279 	VM_REG_GUEST_DS,
280 	VM_REG_GUEST_FS,
281 	VM_REG_GUEST_GS,
282 	VM_REG_GUEST_LDTR,
283 	VM_REG_GUEST_TR,
284 	VM_REG_GUEST_IDTR,
285 	VM_REG_GUEST_GDTR,
286 	VM_REG_GUEST_EFER,
287 	VM_REG_GUEST_CR2,
288 	VM_REG_LAST
289 };
290 
291 /*
292  * Identifiers for optional vmm capabilities
293  */
294 enum vm_cap_type {
295 	VM_CAP_HALT_EXIT,
296 	VM_CAP_MTRAP_EXIT,
297 	VM_CAP_PAUSE_EXIT,
298 	VM_CAP_UNRESTRICTED_GUEST,
299 	VM_CAP_ENABLE_INVPCID,
300 	VM_CAP_MAX
301 };
302 
303 enum x2apic_state {
304 	X2APIC_DISABLED,
305 	X2APIC_ENABLED,
306 	X2APIC_STATE_LAST
307 };
308 
309 enum vm_intr_trigger {
310 	EDGE_TRIGGER,
311 	LEVEL_TRIGGER
312 };
313 
314 /*
315  * The 'access' field has the format specified in Table 21-2 of the Intel
316  * Architecture Manual vol 3b.
317  *
318  * XXX The contents of the 'access' field are architecturally defined except
319  * bit 16 - Segment Unusable.
320  */
321 struct seg_desc {
322 	uint64_t	base;
323 	uint32_t	limit;
324 	uint32_t	access;
325 };
326 #define	SEG_DESC_TYPE(desc)		((desc)->access & 0x001f)
327 #define	SEG_DESC_PRESENT(desc)		((desc)->access & 0x0080)
328 #define	SEG_DESC_DEF32(desc)		((desc)->access & 0x4000)
329 #define	SEG_DESC_GRANULARITY(desc)	((desc)->access & 0x8000)
330 #define	SEG_DESC_UNUSABLE(desc)		((desc)->access & 0x10000)
331 
332 enum vm_cpu_mode {
333 	CPU_MODE_COMPATIBILITY,		/* IA-32E mode (CS.L = 0) */
334 	CPU_MODE_64BIT,			/* IA-32E mode (CS.L = 1) */
335 };
336 
337 enum vm_paging_mode {
338 	PAGING_MODE_FLAT,
339 	PAGING_MODE_32,
340 	PAGING_MODE_PAE,
341 	PAGING_MODE_64,
342 };
343 
344 struct vm_guest_paging {
345 	uint64_t	cr3;
346 	int		cpl;
347 	enum vm_cpu_mode cpu_mode;
348 	enum vm_paging_mode paging_mode;
349 };
350 
351 /*
352  * The data structures 'vie' and 'vie_op' are meant to be opaque to the
353  * consumers of instruction decoding. The only reason why their contents
354  * need to be exposed is because they are part of the 'vm_exit' structure.
355  */
356 struct vie_op {
357 	uint8_t		op_byte;	/* actual opcode byte */
358 	uint8_t		op_type;	/* type of operation (e.g. MOV) */
359 	uint16_t	op_flags;
360 };
361 
362 #define	VIE_INST_SIZE	15
363 struct vie {
364 	uint8_t		inst[VIE_INST_SIZE];	/* instruction bytes */
365 	uint8_t		num_valid;		/* size of the instruction */
366 	uint8_t		num_processed;
367 
368 	uint8_t		rex_w:1,		/* REX prefix */
369 			rex_r:1,
370 			rex_x:1,
371 			rex_b:1,
372 			rex_present:1;
373 
374 	uint8_t		mod:2,			/* ModRM byte */
375 			reg:4,
376 			rm:4;
377 
378 	uint8_t		ss:2,			/* SIB byte */
379 			index:4,
380 			base:4;
381 
382 	uint8_t		disp_bytes;
383 	uint8_t		imm_bytes;
384 
385 	uint8_t		scale;
386 	int		base_register;		/* VM_REG_GUEST_xyz */
387 	int		index_register;		/* VM_REG_GUEST_xyz */
388 
389 	int64_t		displacement;		/* optional addr displacement */
390 	int64_t		immediate;		/* optional immediate operand */
391 
392 	uint8_t		decoded;	/* set to 1 if successfully decoded */
393 
394 	struct vie_op	op;			/* opcode description */
395 };
396 
397 enum vm_exitcode {
398 	VM_EXITCODE_INOUT,
399 	VM_EXITCODE_VMX,
400 	VM_EXITCODE_BOGUS,
401 	VM_EXITCODE_RDMSR,
402 	VM_EXITCODE_WRMSR,
403 	VM_EXITCODE_HLT,
404 	VM_EXITCODE_MTRAP,
405 	VM_EXITCODE_PAUSE,
406 	VM_EXITCODE_PAGING,
407 	VM_EXITCODE_INST_EMUL,
408 	VM_EXITCODE_SPINUP_AP,
409 	VM_EXITCODE_DEPRECATED1,	/* used to be SPINDOWN_CPU */
410 	VM_EXITCODE_RENDEZVOUS,
411 	VM_EXITCODE_IOAPIC_EOI,
412 	VM_EXITCODE_SUSPENDED,
413 	VM_EXITCODE_INOUT_STR,
414 	VM_EXITCODE_MAX
415 };
416 
417 struct vm_inout {
418 	uint16_t	bytes:3;	/* 1 or 2 or 4 */
419 	uint16_t	in:1;
420 	uint16_t	string:1;
421 	uint16_t	rep:1;
422 	uint16_t	port;
423 	uint32_t	eax;		/* valid for out */
424 };
425 
426 struct vm_inout_str {
427 	struct vm_inout	inout;		/* must be the first element */
428 	struct vm_guest_paging paging;
429 	uint64_t	rflags;
430 	uint64_t	cr0;
431 	uint64_t	index;
432 	uint64_t	count;		/* rep=1 (%rcx), rep=0 (1) */
433 	int		addrsize;
434 	enum vm_reg_name seg_name;
435 	struct seg_desc seg_desc;
436 };
437 
438 struct vm_exit {
439 	enum vm_exitcode	exitcode;
440 	int			inst_length;	/* 0 means unknown */
441 	uint64_t		rip;
442 	union {
443 		struct vm_inout	inout;
444 		struct vm_inout_str inout_str;
445 		struct {
446 			uint64_t	gpa;
447 			int		fault_type;
448 		} paging;
449 		struct {
450 			uint64_t	gpa;
451 			uint64_t	gla;
452 			struct vm_guest_paging paging;
453 			struct vie	vie;
454 		} inst_emul;
455 		/*
456 		 * VMX specific payload. Used when there is no "better"
457 		 * exitcode to represent the VM-exit.
458 		 */
459 		struct {
460 			int		status;		/* vmx inst status */
461 			/*
462 			 * 'exit_reason' and 'exit_qualification' are valid
463 			 * only if 'status' is zero.
464 			 */
465 			uint32_t	exit_reason;
466 			uint64_t	exit_qualification;
467 			/*
468 			 * 'inst_error' and 'inst_type' are valid
469 			 * only if 'status' is non-zero.
470 			 */
471 			int		inst_type;
472 			int		inst_error;
473 		} vmx;
474 		struct {
475 			uint32_t	code;		/* ecx value */
476 			uint64_t	wval;
477 		} msr;
478 		struct {
479 			int		vcpu;
480 			uint64_t	rip;
481 		} spinup_ap;
482 		struct {
483 			uint64_t	rflags;
484 		} hlt;
485 		struct {
486 			int		vector;
487 		} ioapic_eoi;
488 		struct {
489 			enum vm_suspend_how how;
490 		} suspended;
491 	} u;
492 };
493 
494 #endif	/* _VMM_H_ */
495